fix: wrong calculation for WeightedMean variance NaN's via UHI #695
Conversation
| where=view.sum_of_weights > 1, # type: ignore[union-attr] | ||
| where=view.sum_of_weights**2 > view.sum_of_weights_squared, |
There was a problem hiding this comment.
What the... Black, this comment was needed!
There was a problem hiding this comment.
Opened an issue at psf/black#2873, and worked around this by moving this subexpression to a variable, which is a bit more readable, too - but it should not have deleted this comment!
There was a problem hiding this comment.
(I do plan to eventually make this properly generic, which would fix most or all of these typing issues - probably will be after we drop 3.6, though)
henryiii
force-pushed
the
henryiii/fix/var
branch
from
9a5500c to
a01e445
Compare
henryiii
force-pushed
the
henryiii/fix/var
branch
from
9f260dc to
1bdca89
Compare
henryiii
force-pushed
the
henryiii/fix/var
branch
from
1bdca89 to
458bcea
Compare
|
I see you already merged this, so I guess I don't have to check it? |
|
You can still check it, it was just a bad bug and the new behavior is at least closer to correct; we can always make a followup PR. Any ideas for further situations to test are always welcome. :) Starting a discussion (#701) was a good idea. |
Brought up on Gitter. The old calculation could not have been right, and was not what is described above from the UHI.